- File:
-
- 1 edited
-
libcfa/src/concurrency/ready_queue.cfa (modified) (24 diffs)
Legend:
- Unmodified
- Added
- Removed
-
libcfa/src/concurrency/ready_queue.cfa
r0fb3ee5 r75c7252 20 20 21 21 22 //#define USE_RELAXED_FIFO22 #define USE_RELAXED_FIFO 23 23 // #define USE_WORK_STEALING 24 24 // #define USE_CPU_WORK_STEALING 25 #define USE_AWARE_STEALING26 25 27 26 #include "bits/defs.hfa" … … 30 29 31 30 #include "stdlib.hfa" 32 #include "limits.hfa"33 31 #include "math.hfa" 34 32 … … 56 54 #endif 57 55 58 #if defined(USE_AWARE_STEALING) 59 #define READYQ_SHARD_FACTOR 2 60 #define SEQUENTIAL_SHARD 2 61 #elif defined(USE_CPU_WORK_STEALING) 56 #if defined(USE_CPU_WORK_STEALING) 62 57 #define READYQ_SHARD_FACTOR 2 63 58 #elif defined(USE_RELAXED_FIFO) … … 143 138 __kernel_rseq_register(); 144 139 140 __cfadbg_print_safe(ready_queue, "Kernel : Registering proc %p for RW-Lock\n", proc); 145 141 bool * handle = (bool *)&kernelTLS().sched_lock; 146 142 … … 178 174 } 179 175 176 __cfadbg_print_safe(ready_queue, "Kernel : Registering proc %p done, id %lu\n", proc, n); 177 180 178 // Return new spot. 181 179 /* paranoid */ verify(n < ready); … … 192 190 193 191 __atomic_store_n(cell, 0p, __ATOMIC_RELEASE); 192 193 __cfadbg_print_safe(ready_queue, "Kernel : Unregister proc %p\n", proc); 194 194 195 195 __kernel_rseq_unregister(); … … 244 244 245 245 //======================================================================= 246 // caches handling247 248 struct __attribute__((aligned(128))) __ready_queue_caches_t {249 // Count States:250 // - 0 : No one is looking after this cache251 // - 1 : No one is looking after this cache, BUT it's not empty252 // - 2+ : At least one processor is looking after this cache253 volatile unsigned count;254 };255 256 void ?{}(__ready_queue_caches_t & this) { this.count = 0; }257 void ^?{}(__ready_queue_caches_t & this) {}258 259 static inline void depart(__ready_queue_caches_t & cache) {260 /* paranoid */ verify( cache.count > 1);261 __atomic_fetch_add(&cache.count, -1, __ATOMIC_SEQ_CST);262 /* paranoid */ verify( cache.count != 0);263 /* paranoid */ verify( cache.count < 65536 ); // This verify assumes no cluster will have more than 65000 kernel threads mapped to a single cache, which could be correct but is super weird.264 }265 266 static inline void arrive(__ready_queue_caches_t & cache) {267 // for() {268 // unsigned expected = cache.count;269 // unsigned desired = 0 == expected ? 2 : expected + 1;270 // }271 }272 273 //=======================================================================274 246 // Cforall Ready Queue used for scheduling 275 247 //======================================================================= 276 unsigned long long moving_average(unsigned long long currtsc, unsigned long long instsc, unsigned long long old_avg) { 277 /* paranoid */ verifyf( currtsc < 45000000000000000, "Suspiciously large current time: %'llu (%llx)\n", currtsc, currtsc ); 278 /* paranoid */ verifyf( instsc < 45000000000000000, "Suspiciously large insert time: %'llu (%llx)\n", instsc, instsc ); 279 /* paranoid */ verifyf( old_avg < 15000000000000, "Suspiciously large previous average: %'llu (%llx)\n", old_avg, old_avg ); 280 281 const unsigned long long new_val = currtsc > instsc ? currtsc - instsc : 0; 282 const unsigned long long total_weight = 16; 283 const unsigned long long new_weight = 4; 284 const unsigned long long old_weight = total_weight - new_weight; 285 const unsigned long long ret = ((new_weight * new_val) + (old_weight * old_avg)) / total_weight; 286 return ret; 248 unsigned long long moving_average(unsigned long long nval, unsigned long long oval) { 249 const unsigned long long tw = 16; 250 const unsigned long long nw = 4; 251 const unsigned long long ow = tw - nw; 252 return ((nw * nval) + (ow * oval)) / tw; 287 253 } 288 254 … … 305 271 } 306 272 #else 307 lanes.data = 0p; 308 lanes.tscs = 0p; 309 lanes.caches = 0p; 310 lanes.help = 0p; 311 lanes.count = 0; 273 lanes.data = 0p; 274 lanes.tscs = 0p; 275 lanes.help = 0p; 276 lanes.count = 0; 312 277 #endif 313 278 } … … 320 285 free(lanes.data); 321 286 free(lanes.tscs); 322 free(lanes.caches);323 287 free(lanes.help); 324 288 } 325 289 326 290 //----------------------------------------------------------------------- 327 #if defined(USE_AWARE_STEALING)328 __attribute__((hot)) void push(struct cluster * cltr, struct thread$ * thrd, unpark_hint hint) with (cltr->ready_queue) {329 processor * const proc = kernelTLS().this_processor;330 const bool external = (!proc) || (cltr != proc->cltr);331 const bool remote = hint == UNPARK_REMOTE;332 333 unsigned i;334 if( external || remote ) {335 // Figure out where thread was last time and make sure it's valid336 /* paranoid */ verify(thrd->preferred >= 0);337 if(thrd->preferred * READYQ_SHARD_FACTOR < lanes.count) {338 /* paranoid */ verify(thrd->preferred * READYQ_SHARD_FACTOR < lanes.count);339 unsigned start = thrd->preferred * READYQ_SHARD_FACTOR;340 do {341 unsigned r = __tls_rand();342 i = start + (r % READYQ_SHARD_FACTOR);343 /* paranoid */ verify( i < lanes.count );344 // If we can't lock it retry345 } while( !__atomic_try_acquire( &lanes.data[i].lock ) );346 } else {347 do {348 i = __tls_rand() % lanes.count;349 } while( !__atomic_try_acquire( &lanes.data[i].lock ) );350 }351 } else {352 do {353 unsigned r = proc->rdq.its++;354 i = proc->rdq.id + (r % READYQ_SHARD_FACTOR);355 /* paranoid */ verify( i < lanes.count );356 // If we can't lock it retry357 } while( !__atomic_try_acquire( &lanes.data[i].lock ) );358 }359 360 // Actually push it361 push(lanes.data[i], thrd);362 363 // Unlock and return364 __atomic_unlock( &lanes.data[i].lock );365 366 #if !defined(__CFA_NO_STATISTICS__)367 if(unlikely(external || remote)) __atomic_fetch_add(&cltr->stats->ready.push.extrn.success, 1, __ATOMIC_RELAXED);368 else __tls_stats()->ready.push.local.success++;369 #endif370 }371 372 static inline unsigned long long calc_cutoff(const unsigned long long ctsc, const processor * proc, __ready_queue_t & rdq) {373 unsigned start = proc->rdq.id;374 unsigned long long max = 0;375 for(i; READYQ_SHARD_FACTOR) {376 unsigned long long ptsc = ts(rdq.lanes.data[start + i]);377 if(ptsc != -1ull) {378 /* paranoid */ verify( start + i < rdq.lanes.count );379 unsigned long long tsc = moving_average(ctsc, ptsc, rdq.lanes.tscs[start + i].ma);380 if(tsc > max) max = tsc;381 }382 }383 return (max + 2 * max) / 2;384 }385 386 __attribute__((hot)) struct thread$ * pop_fast(struct cluster * cltr) with (cltr->ready_queue) {387 /* paranoid */ verify( lanes.count > 0 );388 /* paranoid */ verify( kernelTLS().this_processor );389 /* paranoid */ verify( kernelTLS().this_processor->rdq.id < lanes.count );390 391 processor * const proc = kernelTLS().this_processor;392 unsigned this = proc->rdq.id;393 /* paranoid */ verify( this < lanes.count );394 __cfadbg_print_safe(ready_queue, "Kernel : pop from %u\n", this);395 396 // Figure out the current cpu and make sure it is valid397 const int cpu = __kernel_getcpu();398 /* paranoid */ verify(cpu >= 0);399 /* paranoid */ verify(cpu < cpu_info.hthrd_count);400 unsigned this_cache = cpu_info.llc_map[cpu].cache;401 402 // Super important: don't write the same value over and over again403 // We want to maximise our chances that his particular values stays in cache404 if(lanes.caches[this / READYQ_SHARD_FACTOR].id != this_cache)405 __atomic_store_n(&lanes.caches[this / READYQ_SHARD_FACTOR].id, this_cache, __ATOMIC_RELAXED);406 407 const unsigned long long ctsc = rdtscl();408 409 if(proc->rdq.target == MAX) {410 uint64_t chaos = __tls_rand();411 unsigned ext = chaos & 0xff;412 unsigned other = (chaos >> 8) % (lanes.count);413 414 if(ext < 3 || __atomic_load_n(&lanes.caches[other / READYQ_SHARD_FACTOR].id, __ATOMIC_RELAXED) == this_cache) {415 proc->rdq.target = other;416 }417 }418 else {419 const unsigned target = proc->rdq.target;420 __cfadbg_print_safe(ready_queue, "Kernel : %u considering helping %u, tcsc %llu\n", this, target, lanes.tscs[target].tv);421 /* paranoid */ verify( lanes.tscs[target].tv != MAX );422 if(target < lanes.count) {423 const unsigned long long cutoff = calc_cutoff(ctsc, proc, cltr->ready_queue);424 const unsigned long long age = moving_average(ctsc, lanes.tscs[target].tv, lanes.tscs[target].ma);425 __cfadbg_print_safe(ready_queue, "Kernel : Help attempt on %u from %u, age %'llu vs cutoff %'llu, %s\n", target, this, age, cutoff, age > cutoff ? "yes" : "no");426 if(age > cutoff) {427 thread$ * t = try_pop(cltr, target __STATS(, __tls_stats()->ready.pop.help));428 if(t) return t;429 }430 }431 proc->rdq.target = MAX;432 }433 434 for(READYQ_SHARD_FACTOR) {435 unsigned i = this + (proc->rdq.itr++ % READYQ_SHARD_FACTOR);436 if(thread$ * t = try_pop(cltr, i __STATS(, __tls_stats()->ready.pop.local))) return t;437 }438 439 // All lanes where empty return 0p440 return 0p;441 442 }443 __attribute__((hot)) struct thread$ * pop_slow(struct cluster * cltr) with (cltr->ready_queue) {444 unsigned i = __tls_rand() % lanes.count;445 return try_pop(cltr, i __STATS(, __tls_stats()->ready.pop.steal));446 }447 __attribute__((hot)) struct thread$ * pop_search(struct cluster * cltr) {448 return search(cltr);449 }450 #endif451 291 #if defined(USE_CPU_WORK_STEALING) 452 292 __attribute__((hot)) void push(struct cluster * cltr, struct thread$ * thrd, unpark_hint hint) with (cltr->ready_queue) { … … 510 350 /* paranoid */ verify( kernelTLS().this_processor ); 511 351 512 processor * const proc = kernelTLS().this_processor;513 352 const int cpu = __kernel_getcpu(); 514 353 /* paranoid */ verify(cpu >= 0); … … 521 360 /* paranoid */ verifyf((map.start + map.count) * READYQ_SHARD_FACTOR <= lanes.count, "have %zu lanes but map can go up to %u", lanes.count, (map.start + map.count) * READYQ_SHARD_FACTOR); 522 361 362 processor * const proc = kernelTLS().this_processor; 523 363 const int start = map.self * READYQ_SHARD_FACTOR; 524 364 const unsigned long long ctsc = rdtscl(); 525 365 526 366 // Did we already have a help target 527 if(proc->rdq.target == MAX) {367 if(proc->rdq.target == -1u) { 528 368 unsigned long long max = 0; 529 369 for(i; READYQ_SHARD_FACTOR) { 530 unsigned long long tsc = moving_average(ctsc ,ts(lanes.data[start + i]), lanes.tscs[start + i].ma);370 unsigned long long tsc = moving_average(ctsc - ts(lanes.data[start + i]), lanes.tscs[start + i].ma); 531 371 if(tsc > max) max = tsc; 532 372 } 533 //proc->rdq.cutoff = (max + 2 * max) / 2;373 proc->rdq.cutoff = (max + 2 * max) / 2; 534 374 /* paranoid */ verify(lanes.count < 65536); // The following code assumes max 65536 cores. 535 375 /* paranoid */ verify(map.count < 65536); // The following code assumes max 65536 cores. … … 544 384 } 545 385 546 /* paranoid */ verify(proc->rdq.target != MAX);386 /* paranoid */ verify(proc->rdq.target != -1u); 547 387 } 548 388 else { 549 389 unsigned long long max = 0; 550 390 for(i; READYQ_SHARD_FACTOR) { 551 unsigned long long tsc = moving_average(ctsc ,ts(lanes.data[start + i]), lanes.tscs[start + i].ma);391 unsigned long long tsc = moving_average(ctsc - ts(lanes.data[start + i]), lanes.tscs[start + i].ma); 552 392 if(tsc > max) max = tsc; 553 393 } … … 555 395 { 556 396 unsigned target = proc->rdq.target; 557 proc->rdq.target = MAX;397 proc->rdq.target = -1u; 558 398 lanes.help[target / READYQ_SHARD_FACTOR].tri++; 559 if(moving_average(ctsc ,lanes.tscs[target].tv, lanes.tscs[target].ma) > cutoff) {399 if(moving_average(ctsc - lanes.tscs[target].tv, lanes.tscs[target].ma) > cutoff) { 560 400 thread$ * t = try_pop(cltr, target __STATS(, __tls_stats()->ready.pop.help)); 561 401 proc->rdq.last = target; 562 402 if(t) return t; 403 else proc->rdq.target = -1u; 563 404 } 564 proc->rdq.target = MAX;405 else proc->rdq.target = -1u; 565 406 } 566 407 567 408 unsigned last = proc->rdq.last; 568 if(last != MAX && moving_average(ctsc, lanes.tscs[last].tv, lanes.tscs[last].ma) >cutoff) {409 if(last != -1u && lanes.tscs[last].tv < cutoff && ts(lanes.data[last]) < cutoff) { 569 410 thread$ * t = try_pop(cltr, last __STATS(, __tls_stats()->ready.pop.help)); 570 411 if(t) return t; 571 412 } 572 413 else { 573 proc->rdq.last = MAX;414 proc->rdq.last = -1u; 574 415 } 575 416 } … … 587 428 processor * const proc = kernelTLS().this_processor; 588 429 unsigned last = proc->rdq.last; 589 if(last != MAX) {430 if(last != -1u) { 590 431 struct thread$ * t = try_pop(cltr, last __STATS(, __tls_stats()->ready.pop.steal)); 591 432 if(t) return t; 592 proc->rdq.last = MAX;433 proc->rdq.last = -1u; 593 434 } 594 435 … … 719 560 #else 720 561 unsigned preferred = thrd->preferred; 721 const bool external = (hint != UNPARK_LOCAL) || (!kernelTLS().this_processor) || preferred == MAX|| thrd->curr_cluster != cltr;562 const bool external = (hint != UNPARK_LOCAL) || (!kernelTLS().this_processor) || preferred == -1u || thrd->curr_cluster != cltr; 722 563 /* paranoid */ verifyf(external || preferred < lanes.count, "Invalid preferred queue %u for %u lanes", preferred, lanes.count ); 723 564 … … 771 612 processor * proc = kernelTLS().this_processor; 772 613 773 if(proc->rdq.target == MAX) {614 if(proc->rdq.target == -1u) { 774 615 unsigned long long min = ts(lanes.data[proc->rdq.id]); 775 616 for(int i = 0; i < READYQ_SHARD_FACTOR; i++) { … … 782 623 else { 783 624 unsigned target = proc->rdq.target; 784 proc->rdq.target = MAX;625 proc->rdq.target = -1u; 785 626 const unsigned long long bias = 0; //2_500_000_000; 786 627 const unsigned long long cutoff = proc->rdq.cutoff > bias ? proc->rdq.cutoff - bias : proc->rdq.cutoff; … … 817 658 // try to pop from a lane given by index w 818 659 static inline struct thread$ * try_pop(struct cluster * cltr, unsigned w __STATS(, __stats_readyQ_pop_t & stats)) with (cltr->ready_queue) { 819 /* paranoid */ verify( w < lanes.count );820 660 __STATS( stats.attempt++; ) 821 661 … … 841 681 // Actually pop the list 842 682 struct thread$ * thrd; 843 #if defined(USE_AWARE_STEALING) || defined(USE_WORK_STEALING) || defined(USE_CPU_WORK_STEALING) 844 unsigned long long tsc_before = ts(lane); 845 #endif 683 unsigned long long tsc_before = ts(lane); 846 684 unsigned long long tsv; 847 685 [thrd, tsv] = pop(lane); … … 857 695 __STATS( stats.success++; ) 858 696 859 #if defined(USE_AWARE_STEALING) || defined(USE_WORK_STEALING) || defined(USE_CPU_WORK_STEALING) 860 if (tsv != MAX) { 861 unsigned long long now = rdtscl(); 862 unsigned long long pma = __atomic_load_n(&lanes.tscs[w].ma, __ATOMIC_RELAXED); 863 __atomic_store_n(&lanes.tscs[w].tv, tsv, __ATOMIC_RELAXED); 864 __atomic_store_n(&lanes.tscs[w].ma, moving_average(now, tsc_before, pma), __ATOMIC_RELAXED); 865 } 697 #if defined(USE_WORK_STEALING) || defined(USE_CPU_WORK_STEALING) 698 unsigned long long now = rdtscl(); 699 lanes.tscs[w].tv = tsv; 700 lanes.tscs[w].ma = moving_average(now > tsc_before ? now - tsc_before : 0, lanes.tscs[w].ma); 866 701 #endif 867 702 868 #if defined(USE_ AWARE_STEALING) || defined(USE_CPU_WORK_STEALING)703 #if defined(USE_CPU_WORK_STEALING) 869 704 thrd->preferred = w / READYQ_SHARD_FACTOR; 870 705 #else … … 965 800 /* paranoid */ verifyf( it, "Unexpected null iterator, at index %u of %u\n", i, count); 966 801 it->rdq.id = value; 967 it->rdq.target = MAX;802 it->rdq.target = -1u; 968 803 value += READYQ_SHARD_FACTOR; 969 804 it = &(*it)`next; … … 978 813 979 814 static void fix_times( struct cluster * cltr ) with( cltr->ready_queue ) { 980 #if defined(USE_ AWARE_STEALING) || defined(USE_WORK_STEALING)815 #if defined(USE_WORK_STEALING) 981 816 lanes.tscs = alloc(lanes.count, lanes.tscs`realloc); 982 817 for(i; lanes.count) { 983 lanes.tscs[i].tv = rdtscl(); 984 lanes.tscs[i].ma = 0; 818 unsigned long long tsc1 = ts(lanes.data[i]); 819 unsigned long long tsc2 = rdtscl(); 820 lanes.tscs[i].tv = min(tsc1, tsc2); 985 821 } 986 822 #endif … … 1028 864 // Update original 1029 865 lanes.count = ncount; 1030 1031 lanes.caches = alloc( target, lanes.caches`realloc );1032 866 } 1033 867 … … 1106 940 fix(lanes.data[idx]); 1107 941 } 1108 1109 lanes.caches = alloc( target, lanes.caches`realloc );1110 942 } 1111 943 1112 944 fix_times(cltr); 1113 1114 945 1115 946 reassign_cltr_id(cltr);
Note:
See TracChangeset
for help on using the changeset viewer.
